Fig 1.
Overview of the sampling design.
Three major spatio-temporal dimensions were considered for arthropods: spatial turnover or difference in species composition measured (A) horizontally (among sites, all less than 2km apart), (B) vertically from -5cm to 36m above ground, and (C) temporally, among sampling intervals repeated within a period of 413 days. (A) Twelve 20x20m sites (I-XII) were surveyed for plants and arthropods, from the ground to the upper canopy. (B) Arthopods were surveyed using 14 different protocols from ground to the upper canopy: (1) baits and netting; (2) gall sampling within the volumetric space of a vertical cylinder; (3) sticky traps; (4) aerial composite flight-interception traps; (5) beating of vegetation and dead branches; (6) hand collecting for ants and termites; (7) ground flight-interception traps; (8) collection of the leaf-litter fauna and extraction with a mini-Winkler apparatus; (9) collection of ground and suspended soils, extraction with Berlese-Tullgren apparatus; (10) wood rearing; (11) pitfall traps; (12) ground Malaise traps; (13) canopy fogging; (14) light traps. (C) After an initial sampling period of 6 weeks during the late wet season (September-October 2003, hereafter Survey S1), field work was replicated during three similar sampling periods targeting the dry, early wet and late wet seasons (Survey S2: February-March 2004, Survey S3: May-June 2004, Survey S4: October-November 2004). Photos by JS (1), SR (2), ML (3,5,8,9,11,13), NW (4), N. Baiben (6), C.E. Carlton (7), M. Janda & J. Patera (10), RLK (14), S. Pinzon (12).
Table 1.
Results of Kruskall-Wallis tests comparing arthropod abundance among sites, habitats and surveys.
Too few samples were available for a composite analysis of habitats.
Fig 2.
Representative box-plots of arthropod abundance across (a) sites, (b) habitats and (c) surveys.
Fig 3.
Mean (± s.e.) abundance per sample, detailed per arthropod guild and habitats (black bars = litter, grey bars = understory, stippled bars = canopy, white bars = upper canopy).
ANOVAs comparing mean abundance per habitat within guilds are all significant with at least p<0.01. For each guild, different letters denote significantly different means (Tukey tests, p<0.05).
Fig 4.
Additive decomposition of species richness for (a) major data sets and (b) arthropod guilds.
(a) Major data sets (estimated species richness with ten protocols, observed species richness with ten protocols, estimated species richness with FITs, observed species richness with FITs, 885 common and rare species, inset: flowering trees (note the different scale). (b) Arthropod guilds (observed species richness with ten protocols). Species richness is partitioned among alpha (black bars), betaT (grey bars), betaH (stippled bars) and betaV (white bars). * indicates that parameters are significantly different from zero. Randomization tests were not performed with estimated species richness.
Fig 5.
Venn diagrams summarizing canonical variation partitioning along three dimensions for (a) arthropods and (b) flowering trees.
(a) 5858 arthropod species collected at 12 sites with all collecting methods. (b) 25 species of trees flowering during Surveys 1–4. Variables used to characterize the three dimensions are included in grey boxes. ** = significance of the fraction of variation (200 randomizations, p< 0.01); # fraction not testable. For description of fractions [a]-[h], see methods.
Fig 6.
Arthropod species turnover, expressed by faunal similarity measured with the Morisita-Horn index, in the (a) horizontal, (b) vertical and (c) seasonal dimensions.
Shown are the parameters of pairwise dissimilarity regressed on pairwise log(distance), with p values based on 1,000 permutations of pairwise distance versus pairwise dissimilarity matrices, and the overall concordance (r) between the matrices of observed and estimated values. Plotted models refer to the decay of similarity (i.e. 1-dissimilarity), for more intuitive interpretation.
Table 2.
Summary of the principal results, detailed for the horizontal, vertical and seasonal dimensions.
0 = no interaction, + = weak, ++ = intermediate, +++ = strong, NA = not available (see results), NT not tested (see methods). AR = arthropods, TR = trees.